Machine learning optimal ordering in global routing problems in semiconductors.

Area of Science:

• Computer Science
• Electrical Engineering
• Artificial Intelligence

Background:

• Global routing is a critical step in designing multilayered semiconductor packages.
• Current net ordering methods often rely on heuristic score functions, which can be suboptimal.
• Efficient net ordering is essential for minimizing routing congestion and improving signal integrity.

Purpose of the Study:

• To propose a new machine learning-based method for net ordering in global routing.
• To demonstrate the superiority of the proposed method over conventional heuristic techniques.
• To enhance the layer assignment process in multilayered semiconductor package design.

Main Methods:

• Development of a novel machine learning algorithm for net ordering.
• Application of deep learning techniques to the net ordering problem.
• Comparative analysis through global routing experiments in multilayered semiconductor package environments.

Main Results:

• The proposed machine learning method significantly outperforms traditional heuristic net ordering techniques.
• Experimental results validate the effectiveness of the new approach in global routing.
• The net ordering step in global routing can be substantially improved using deep learning.

Conclusions:

• Machine learning, particularly deep learning, offers a powerful alternative for net ordering in global routing.
• The proposed method provides a more effective solution for routing problems in semiconductor package design.
• This research advances the application of AI in optimizing complex electronic design automation tasks.